www.computerweekly.com
Labour has published its 'AI Opportunities Action Plan' (The Plan). The Prime Minister is very bullish about The Plan, and peppers his foreword with muscular terms like growth, revolution, ambition, strength and innovation. In itself, The Plan is full of claims that AI is essential and inevitable, and urges the government to pour public money into the industry so as not to miss out.In the style of tech entrepreneurs, The Plan likes to put 'x' after things, so investment must go up by 20x (meaning twenty times), the amount of compute AI requires has already gone up by 10,000x and so on. The Plan claims that Britain is already leading the world through the AI Safety Institute (of which more later) and infuses the usual AI hype with nationalist vibes via terms like world leader, world-class, national champions and 'Sovereign AI'.Above all, The Plan emphasises the need to scale. The significance of scale for AI and its technopolitical impacts will be explored below.This article addresses the poverty of The Plan and the emptiness of its claims about AI but, rather than a point-by-point rebuttal, it's about the underlying reasons why this Labour government supports measures that will harm both people and the environment.In between invocations of speed, pace and scale, there's some recognition in The Plan that the UK is not a wholly happy place right now. While recommending a high-tech form of land enclosures via 'AI Growth Zones' (AIGZs), which are about handing data centre developers "access to land and power", it gestures towards the idea that these could drive local innovation in post-industrial towns. While The Plan's claims about AI's inevitable progress and the oncoming wave of agentic systems that will reason, plan and act for themselves already seem dated and discredited, what hasn't changed is that the very regions targeted for growth via AIGZs have already seen violent anti-immigrant pogroms accompanied by fascist rhetoric, and those sentiments have not gone away.Ultimately, it's argued here, the misstep represented by The Plan and its total commitment to AI will reinforce and amplify the threat of the far right, as well as connecting it to the extremely reactionary ideas that are in the ascendency in Silicon Valley. This article proposes instead 'decomputing'; the refusal of hyperscale computing, the replacement of reductive solutionism with structures of care, and the construction of alternative infrastructures based on convivial technologies.In some ways, it's fairly obvious why this Labour government would want to prioritise AI. Kier Starmer's single identifiable political belief is the idea of 'growth', so demonstrating economic growth supersedes all other government concerns.Growth will demonstrate that Kier and colleagues are serious politicians who are no threat to the establishment, and at the same time win over Mr.& Mrs Red Wall voter who are disillusioned with orthodox politics. Boosting GDP is Labour's answer to all the wicked problems that beset the UK's public services and infrastructure, and avoids having to actually challenge the underlying logic of Thatcherite neoliberalism which has dominated for decades.And if there's one area of economic activity which is growing, it's certainly AI. All the graphs are going up; the venture capitalist investment, the stock market valuations, the size of the AI models and the number and scale of the data centres. Latching on to this growth is already working for the government, as a full 10% of the 63b promised by their 'record-breaking' international investment summit was earmarked for data centres.Having apparently decided to pin their hopes on AI, the Labour government have been aligning with demands from Big Tech. Not long after the election that brought Labour to power, Google issued a report titled 'Unlocking the UK's AI Potential' laying out their conditions for AI growth in the UK, including investment in data centres and a loosening of copyright restrictions. Of course, these aren't Google-specific requirements; the foundation of all contemporary AI is scale, meaning more and larger data centres to house all the servers and bigger pools of data to train them.The collision course with copyright comes from the fact that these data sets have always been too large to pay for, so the AI industry just grabs them from the internet without asking. Google's report was accompanied by a media round from their UK & Ireland vice president warning that the UK risked "losing leadership" and "falling behind" if their advice wasn't followed.It seems the message was received loud and clear; since the election, Labour have designated data centres as 'critical national infrastructure' which means that ministers can override any local planning objections, and the government is also floating a relaxation of copyright protections. It's not just Google that the Labour government is prepared to doff its cap to; Peter Kyle, the current secretary of state for Science, Innovation and Technology, has repeatedly stated that the UK should deal with Big Tech via 'statecraft'; in other words, rather than treating AI companies like any other business that needs taxing and regulating, the government should treat the relationships as a matter of diplomatic liaison, as if these entities were on a par with the UK state.This awe of Big Tech reflects deeper currents of commitment within the Labour government. Certainly, any ministers from the Blairite faction are going to be influenced by the absolute belief in AI expressed by influential think tank, the Tony Blair Institute (TBI).It's hard to exaggerate the centrality of AI to the TBI world view, but the title of their 2023 report is pretty representative: 'A New National Purpose: AI Promises a World-Leading Future of Britain'. According to this report, becoming a world leader in AI development is "a matter becoming so urgent and important that how we respond is more likely than anything else to determine Britain's future", and its 2024 sequel, 'Governing in the Age of AI: A New Model to Transform the State', opens with "There is little doubt that AI will change the course of human progress."The breathless rhetoric is accompanied by policy demands; a factor of ten increase in the UK's compute capacity, the diversion of major spending commitments to AI infrastructure, reducing regulation to US levels and, of course, enacting all this in close relationship with the private sector.A core promise is that turning the public sector over to AI will deliver huge savings and improved delivery, although one might question the reliability of their research, given that it was based on asking ChatGPT itself how many government jobs it could do. While this sketchy approach has echoes of the Iraq ('dodgy') Dossier, it's reflecting a realpolitik that sees both AI companies and rhetoric about AI as incredibly powerful at the current moment.This is perhaps the hole that AI fills for the Labour government; having long abandoned any substantive belief in the transformative power of socialism, it is lacking a mobilising belief system. At the same time, it's obvious to all and sundry that the status quo is in deep trouble and that being the party of continuity isn't going to convince anyone.Ergo, the claim that AI has the power to change the world becomes a good stand-in for a transformative ideology. The bonus for the Labour government is that relying on AI to fix things avoids the need for any structural changes that might upset powerful business and media interests, and rhetoric about global AI leadership has a suitably 'Empire' vibe to appeal to nationalistic sentiments at the grassroots.Beneath all the policy gloss and think tank reports, though, lurk the real harms of AI in the here-and-now, starting with environmental harms. The Labour government's vision for AI takes concrete form in the shape of more data centres. However, as some previously tranquil localities are starting to discover, this comes with significant impacts.Generative AI, in particular, is driving the computational scale of AI models through the roof. The rate at which these models are increasing in size outpaces any other recent tech revolution, from smartphone adoption to genome sequencing. In turn, this is driving massive increases in energy demand.To service AI and the internet cloud, the fastest growing type of data centre in the UK is the so-called hyperscale data centres run by the likes of Google, Microsoft and AWS. These are typically at least 10,000sq ft and contain upwards of 5,000 servers, but the industry wants them to be much larger, and filled with the energy-guzzling GPU chips that train and run AI.Sam Altman, CEO of ChatGPT's parent company OpenAI, has pitched plans for 5GW data centres in the US, which is the equivalent of about five nuclear reactors' worth and enough energy to power a large city. These voracious demands for electricity come with immediate consequences for national grids and for climate emissions.The Greater London Authority (GLA) has already had to impose a temporary ban on new housing developments in West London because a cluster of existing data centres was using the available grid supply. Because AI's energy demands are outpacing the development of bigger electricity grids, there's currently a push for bringing back fossil fuel sources, especially gas-powered turbines. Connecting directly to the natural gas network to overcome local power constraints is less efficient than grid-scale generation and increases unmonitored carbon emissions.Of course, Big Tech is already aware that driving climate emissions is a bad look and has previously tried to look 'green' via the use of 'renewables' and the cover story of carbon offsets. However, the scale of generative AI has blown this away to the point where both Google and Microsoft have admitted an inability to meet their own climate targets.Locally, it's not just potential power cuts that an AI data centre brings to an area, but a huge demand for cooling water to stop all the servers overheating and the pervasive presence of a background hum from all the cooling systems. The question is whether a pursuit of 'AI greatness' will make the UK more like Ireland, which has already been recolonised as a dumping ground for Big Tech's data centre infrastructure.The here-and-now harms of AI are also social. Never mind the sci-fi fantasies about AI taking over the world, the mundane reality of AI in any social context are forms of ugly solutionism that perpetuate harms rather than reducing them. The claim that more computation will improve public services is hardly new, and algorithmic fixes for everything from welfare to education have already left a trail of damage in their wake.In Australia, the 'Robodebt' algorithm wrongly accused tens of thousands of people of welfare fraud, and was only halted by a grassroots campaign and an eventual public inquiry, while in the Netherlands an algorithm falsely labelled tens of thousand of people as defrauding the child benefits system, causing crippling debts and family break-ups. What the UK's notorious Horizon IT system and contemporary AI have in common is the tendency to generate falsehoods while appearing to be working properly. What AI adds is the capacity to scale harms in way that makes the Horizon scandal look like small beer.The insistence that AI will reverse the rot in education and healthcare systems also has a tired history. Back in 2018, Facebook's non-profit arm inserted an online learning platform into a California public school system on the basis that it offered 'personalised learning', the central mantra of all AI-driven educational technology. It took mass resistance by 17-year old students to get rid of it.In the open letter they sent to Zuckerberg they said "Most importantly, the entire program eliminates much of the human interaction, teacher support, and discussion and debate with our peers that we need in order to improve our critical thinking. Unlike the claims made in your promotional materials, we students find that we are learning very little to nothing. It's severely damaged our education, and that's why we walked out in protest".Meanwhile the Nobel Prize-winning godfather of AI, Geoffrey Hinton, made the claim back in 2016 that thanks to the superior accuracy of AI's image classification there was no need to train any more radiologists. As it turned out, of course, that claim was just as specious as the more recent hype about ChatGPT passing medical exams. Labour's minister for Science, Innovation and Technology is continuing to use the trope of an AI-powered solution to cancer detection to push for more AI in public services while ignoring calls by leading cancer specialists to "concentrate on the basics of cancer treatment rather than the 'magic bullets' of novel technologies and artificial intelligence".Forcing AI into services in lieu of fixing underlying issues like decaying buildings and without funding more actual teachers and doctors is a form of structural violence - a form of violence by which institutions or social structures harm people through preventing them from meeting their fundamental needs.The political continuity here is that a commitment to AI solutions also enacts a kind of Thatcherite 'shock doctrine' where the sense of urgency generated by an allegedly world-transforming technology is used as an opportunity to transfer power to the private sector.The amount of data and computing power required to create on of today's foundation models, the big generative and supposedly general purpose systems, is beyond the reach of all but the biggest tech companies. Whether it's in welfare, education or health, a shift to AI is a form of privatisation by the backdoor, shifting a significant locus of control to Silicon Valley.Like the original Thatcherism, this is also going to be accompanied by job losses and a change to more precarious forms of algorithm-driven outsourcing. It's Deliveroo all round, not because AI can actually replace jobs but because its shoddy emulations provide managers and employers with a means to pare away employment protections.What marks the Labour government out from earlier forms of neoliberalism is its emphasis on total mobilisation, that all material and human resources should be aligned behind the national mission for growth. This translates into a rhetorical and legislative intolerance for the idea that people should be 'idle', no matter their states of mental distress or other disability.Unfortunately, while AI systems are themselves unproductive in a practical sense, they excel at exactly the functions of ranking, classification and exclusion that are required for forms of social filtering at scale. Turning more services over to AI-assisted decision-making will indeed facilitate the differentiation of the deserving from the undeserving in line with this productivist ideology.This alignment of social and technical ordering will show that Labour is indeed still the 'party of the workers', but only in the sense of squeezing the last drop of work out of people while using algorithmic optimisation to decide who is relatively disposable.Ultimately, the Labour government's capitulation to AI in the vain hope of meaningful growth is a gift to the far right.Most governing parties across Europe are making the mistake of incubating the far right while complaining about their seemingly inexorable rise. Governments seem oblivious to the political fact that trying to distract people with the spectre of the far right while completely failing to address the structural failings of neoliberalism that leave people feeling angry and abandoned, only serves to empower polarising and post-truth politics. Moreover, the fact that populist rhetoric gains support leads the same governing parties to mainstream their reactionary narratives, as the Labour government has done around immigration and the so-called 'small boats crisis'.Using AI to distract from structural problems while failing to deliver actual solutions follows a similar pattern. The only thing these algorithms will do is filter and classify groups of people to blame for the way AI itself degrades the already shoddy state of public services. The double whammy that comes with AI is the way that the industry itself is a catalyst for more extreme right ideologies.The seeds of this can be seen in the apparently innocuous turn to 'AI safety', which was initiated by Rishi Sunak but has been endorsed and continued by the current government. The rationale of AI safety is not to protect people from the everyday harms of dysfunctional AI in their everyday lives, but to head off the imagined potential for AI to trigger human extinction by developing bioweapons or by becoming completely autonomous and simply taking over.This, in turn, derives from the underlying belief in AGI or Artificial General Intelligence; the belief that the humungous but stumbling AI of today is a step to superintelligent systems that will be superior to humans. As Hollywood as this might seem, it's the position of many in AI, from godfather Geoffrey Hinton to the founders of most of the main companies like DeepMind and OpenAI.Such powerfully warping beliefs spawn real world consequences because, ultimately, it's built on a eugenicist mindset. The very idea of 'general intelligence' comes from Victorian eugenics, when scientists like Francis Galton and Karl Pearson were rationalising the racial supremacy that legitimised the British Empire. The idea of superior intelligence always comes with its corollary of inferior intelligence, whether that's defined racially or in terms of disabilities, and always pans out as assessing some lives as more worthy than others.Part of the motive for a belief in AGI is self-serving. If superintelligent AI is our only hope to solve climate change then we shouldn't be limiting the development of AI through small minded measures like carbon emissions targets, but should be mobilising all available resources, fossil fuel or not, behind its accelerated development.This is also good news for fossil fuel oligarchs who, not coincidentally, are some of the biggest funders of far right think tanks.Similarly, if the future of humanity is to join such superintelligence inside computers themselves, and this leads to vastly multiplied numbers of virtual humans, then facilitating the emergence of AGI and the 1054 virtual future humans becomes morally more important than any collateral harms to actual humans in the present moment. Again, while this might seem deranged, it's the stance of a set of beliefs known variously as 'effective altruism' (EA) or 'long termism' which are very influential in Silicon Valley.As a world view, it elevates the self-styled mission of people in AI above the meagre concerns of ordinary folk. Disturbingly, it seems that the infiltration of US and UK policy circles by people with EA beliefs was responsible for the shift to an AI Safety agenda and, in the UK, the creation of an AI Safety Institute.Lurking behind long termism, but equally influential in Silicon Valley, are the darker and more explicitly fascist beliefs of neoreaction . These kinds of ideas, as espoused by the likes of Peter Thiel, argue that democracy is a failed system, that corporations should take over social governance with monarchical models, and that society should optimise itself by weeding out inferior and unproductive elements. It's this strand of far right tech accelerationism that merged with the MAGA movement in the run up to Trump's 2024 re-election.While Elon Musk's support for anti-immigrant pogroms and his direct attacks on Starmer et al have been the most visible consequences for the UK so far, the real threat is the underlying convergence of Big Tech and reactionary politics. AI is not simply a technology but a form of technopolitics, where technology and politics produce and reinforce each other, and in the case of contemporary AI this technopolitics tends towards far right solutionism.Neither Labour nor any other political party is going to defend us against this technopolitics. We can, however, oppose it directly through 'decomputing'.Decomputing starts with the refusal of more data centres for AI, on the basis that they are environmentally damaging and because they run software that's socially harmful. Hyperscale data centres are the platform for AI's assaults on workers' rights, through precaritisation and fake automation, but also for the wider social degradations of everything from preemptive welfare cuts to non-consensual deepfake porn.Decomputing opposes AI because it's built on layers of exploitative labour, much of which is outsourced to the Global South, and because its reductive predictions are foreclosing life chances wherever they're applied to assess our future 'value'.What's needed to salve pain and suffering isn't the enclosure of resources to power the judgements of GPUs but acts of care, the prioritisation of relationships that acknowledge our vulnerabilities and interdependencies.Decomputing is a direct opposition to the material, financial, institutional and conceptual infrastructures of AI not only because they promote an already-failed solutionism but because they massively scale alienation. By injecting even more distancing, abstraction and opacity into our lives, AI is helping to fuel our contemporary crisis, furthering the bitter resentments that feed the far right and the disenchantments that separates us from the more-than-human lifeworld.What we urgently need, instead of a political leadership in thrall to AI's aura of total power, is a reassertion of context and agency by returning control to more local and directly democratic structures. Decomputing argues that, wherever AI is proposed as 'the answer', there is a gap for the self-organisation of people who already know better what needs to be done, whether it's teachers and students resisting generative AI in the classroom or healthcare workers and patients challenging the algorithmic optimisation of workloads that eliminates even minimal chances to relate as human beings.Decomputing claims that the act of opposing AI's intensification of social and environmental harms is at the same time the assertion that other worlds are possible. It parallels contemporary calls for degrowth, which also opposes runaway extractivism by focusing on the alternative structures that could replace it.As much as contemporary AI is a convergence of off-the-scale technology and emergent totalitarianism, decomputing offers a counter-convergence of social movements that brings together issues of workers' rights, feminism, ecology, anti-fascism and international solidarity.Where AI is another iteration in the infrastructuring of Empire, decomputing recognises the urgency of starting to develop alternative infrastructures in the here-and-now, from renewable energy coops to structures of social care based on mutual aid.The murmurings in the financial pages of mainstream media that AI's infinite growth is actually a bubble prone to collapse misses the point that a wider collapse is already upon us in one form or another. Climate change is happening in front of our eyes, while it's pretty clear that liberal democracy is allowing itself to be eaten from the inside by the far right.The more that AI is allowed scale its reactionary technopolitics, the more it will have the effect of narrowing options for the rest of us. Decomputing is the bottom-up recovery of alternatives that have been long buried under techno-fantasies and decades of neoliberalism; people-powered visions of convivial technologies, cooperative socialities and a reassertion of the commons.Read more about artificial intelligenceAI interview: Thomas Dekeyser, researcher and film director: On the politics of 'techno-refusal', and the lessons that can be learned from a clandestine group of French IT workers who spent the early 1980s sabotaging technological infrastructure.AI firms can't be trusted to voluntarily share risk information: Workers at frontier AI firms have warned that their employers including OpenAI, DeepMind and Anthropic can't be trusted to voluntarily share information about their systems capabilities and risks with governments or civil society.Autonomous weapons reduce moral agency and devalue human life: Military technology experts gathered in Vienna have warned about the detrimental psychological effects of AI-powered weapons, arguing that implementing systems of algorithmic-enabled killing dehumanises both the user and the target.
